Predictive Modeling of US Stock Market and Commodities: Impact of Economic Indicators and Geopolitical Events Using Machine

Abstract

Comprehending the dynamics of the stock market is pivotal for investors, policymakers, and economists in the USA. The United States stock market being one of the most influential financial markets in the world, contributes significantly to shaping the world economy. This research project aimed to bridge the important gaps in understanding the interrelationship between economic indicators and geopolitical events concerning the performance of the US stock market and commodities. The prime objective was to assess how different economic indicators have an impact on stock market performance over a specific period. In this regard, the development of machine learning models facilitated the ability to forecast stock market and commodities trends. These models utilized economic indicators and geopolitical events from historical data to predict future movements in the market with higher accuracy than the traditional forecasting technique followed. The study considered several different datasets to comprehensively analyze the effects that economic indicators and geopolitical events have on the stock market and commodity performances. The key datasets used in this analysis involve historical stock market indices such as the S&P 500, Dow Jones Industrial Average, and NASDAQ, as well as commodity prices for gold, oil, and silver. These datasets were collected from reputed financial databases such as Bloomberg and Federal Reserve Economic Data-FRED, for metrics including GDP growth rates, unemployment rates, inflation figures, and interest rates. The researchers also gathered data on geopolitical events: elections, trade wars, and military conflicts, using usually reliable news archives like Reuters, Bloomberg News, and The New York Times. Linear Regression, Random Forest, and XG-Boost algorithms were selected to capture various facets of the data. The performance metrics used to evaluate the models in this study included Mean Absolute Error, Root Mean Squared Error, and R-squared.  Random Forest Regressor outperformed the other models with the lowest RMSE, showcasing its ability to model complex relationships. XG-Boost Regressor equally delivered a strong balance between performance and computational efficiency, achieving similar accuracy to Random Forest. Results from this study therefore can inform policy makers, especially in regards to stabilizing the financial markets during periods of heightened economic or geopolitical uncertainty. Key recommendations include developing proactive policy measures that could dampen the effect of adverse economic indicators and geopolitical events on market stability.